risk aversion and wealth: evidence from person-to-person ... · tal subjects along important...
Post on 22-Mar-2021
1 Views
Preview:
TRANSCRIPT
Risk Aversion and Wealth:
Evidence from Person-to-Person Lending Portfolios ∗
Daniel Paravisini Veronica Rappoport Enrichetta Ravina
LSE, BREAD, CEPR LSE, CEP, CEPR Columbia GSB
August 10, 2015
Abstract
We estimate risk aversion from investors’ financial decisions in a person-to-person lending plat-form. We develop a method that obtains a risk aversion parameter from each portfolio choice.Since the same individuals invest repeatedly, we construct a panel dataset that we use to disen-tangle heterogeneity in attitudes towards risk across investors, from the elasticity of risk aversionto changes in wealth. We find that wealthier investors are more risk averse in the cross section,and that investors become more risk averse after a negative housing wealth shock. Thus, in-vestors exhibit preferences consistent with decreasing relative risk aversion and habit formation.JEL codes: E21, G11, D12, D14.
∗We are grateful to Lending Club for providing the data and for helpful discussions on this project. We thankMichael Adler, Manuel Arellano, Nick Barberis, Geert Bekaert, Patrick Bolton, John Campbell, Marco di Maggio,Larry Glosten, Nagpurnanand Prabhala, Bernard Salanie, and seminar participants at CEMFI, Columbia UniversityGSB, Duke Fuqua School of Business, Hebrew University, Harvard Business School, Kellogg School of Manage-ment, Kellstadt Graduate School of Business-DePaul, London Business School, Maryland Smith School of Busi-ness, M.I.T. Sloan, Universidad Nova de Lisboa, the Yale 2010 Behavioral Science Conference, and the SED 2010meeting for helpful comments. We thank the Program for Financial Studies for financial support. All remainingerrors are our own. Please send correspondence to Daniel Paravisini (d.paravisini@lse.ac.uk), Veronica Rappoport(v.e.rappoport@lse.ac.uk), and Enrichetta Ravina (er2463@columbia.edu).
1 Introduction
Theoretical predictions on investment, asset prices, and the cost of business cycles depend crucially
on assumptions about the relationship between risk aversion and wealth.1 Although characterizing
this relationship has long been in the research agenda of empirical finance and economics, progress
has been hindered by the difficulty of disentangling the shape of the utility function from preference
heterogeneity across agents. For example, the bulk of existing work is based on comparisons of risk
aversion across investors of different wealth, which requires assuming that agents have the same
preference function. If agents have heterogeneous preferences, however, cross sectional analysis
leads to incorrect inferences about the shape of the utility function when wealth and preferences
are correlated. Such a correlation may arise, for example, if agents with heterogeneous propensity to
take risk make different investment choices, which in turn affect their wealth.2 To both characterize
the properties of the joint distribution of preferences and wealth in the cross section, and estimate
the parameters describing the utility function, one needs to observe how the risk aversion of the
same individual changes with wealth shocks.
Important recent work improves on the cross-sectional approach by looking at changes in the
fraction of risky assets in an investor’s portfolio stemming from the time series variation in investor
wealth.3 The crucial identifying assumption required for using the share of risky assets as a proxy for
investor Relative Risk Aversion (RRA) in this setting is that all other determinants of the share of
risky assets remain constant as the investor’s wealth changes. One must assume, for example, that
changes in financial wealth resulting from the performance of risky assets are uncorrelated with
changes in beliefs about their expected return and risk. Attempts to address this identification
problem through an instrumental variable approach have produced mixed results: the estimated
sign of the elasticity of RRA to wealth varies across studies depending on the choice of instrument.4
1 See Kocherlakota (1996) for a discussion of the literature aiming at resolving the equity premium and low risk freerate puzzles under different preference assumptions. Following the seminal contribution in Campbell and Cochrane(1999), recent work shows that preferences with habit formation produce cyclical variations in risk aversion anddecreasing relative risk aversion after a positive wealth shock that can explain these and other empirical asset pricingregularities (Menzly et al., 2004, Buraschi and Jiltsov, 2007, Polkovnichenko, 2007, Yogo, 2008, Korniotis, 2008, andSantos and Veronesi, 2010). Wealth inequality has also been shown to raise the equity premium if the absolute riskaversion is concave in wealth (Gollier, 2001).
2Guvenen (2009) and Gomes and Michaelides (2008) propose a model with preference heterogeneity that endoge-nously generates cross sectional variation in wealth. Alternatively, an unobserved investor characteristic, such ashaving more educated parents, may jointly affect wealth and the propensity to take risk.
3Chiappori and Paiella (2011), Brunnermeier and Nagel (2008), and Calvet et al. (2009).4See Calvet et al. (2009) and Calvet and Sodini (2014) for a discussion.
2
Finally, estimates of risk aversion based on the share of risky assets pose an additional problem
when used to analyze the link between heterogeneity in risk preferences and wealth: measurement
error in wealth is inherited by the risk aversion estimates, and may induce a spurious correlation
between the two. This potentially explains why existing empirical evidence on the correlation
between risk preferences and wealth in the cross section is inconclusive, depends on the definition
of wealth, and is sensitive to the categorization of assets into the risky and riskless categories.5
The present paper exploits a novel environment to obtain, independently of wealth, unbiased
measures of investor risk aversion and to examine the link between these estimates and investor
wealth. We analyze the risk taking behavior of 2,372 investors based on their actual financial
decisions in Lending Club (LC), a person-to-person lending platform in which individuals invest in
diversified portfolios of small loans.6 We develop a methodology to estimate the local curvature of
an investor’s utility function (Absolute Risk Aversion, or ARA) from each portfolio choice. The key
advantage of this estimation approach is that it does not require characterizing investors’ outside
wealth. We also exploit the fact that the same individuals make repeated investments in LC to
construct a panel of risk aversion estimates. We use this panel to both characterize the cross
sectional correlation between risk preferences and wealth, and to obtain reduced form estimates
of the elasticity of investor-specific risk aversion to changes in wealth. We find the cross-sectional
elasticity between Relative Risk Aversion (RRA) and wealth to be positive, and the investor-specific
wealth elasticities to be negative.
Our estimation method is derived from an optimal portfolio model where investors hold an
unknown number of risky and riskless assets, including a portfolio of LC loans. We decompose the
LC returns into a systematic component and an idiosyncratic return component orthogonal to the
market. We use the idiosyncratic component to characterize investors’ preferences: an investor’s
ARA is given by the additional expected return that makes her indifferent about allocating the
marginal dollar to a loan with higher idiosyncratic default probability. Estimating risk preferences
from the non-systematic component of returns implies that the estimates are independent from
the investors’ overall risk exposure or wealth. Moreover, by measuring the curvature of the utility
function directly from the first order condition of this portfolio choice problem, we do not need to
5See, among others, Blume and Friend (1975), Cohn et al. (1975), Morin and Suarez (1983), and Blake (1996).6For prior research analyzing investors’ behavior in peer-to-peer lending see Iyer et al. (2014) , Lin et al. (2013),
Marom and Sade (2013), Ravina (2012).
3
impose a specific shape of the utility function.
This method relies on two assumptions that we validate in the data. The first assumption is
that LC loans are not held to simply replicate the market portfolio. This assumption is validated
by the substantial heterogeneity of mean and variance across investors’ LC portfolios. The second
assumption is that we can correctly capture investors’ beliefs about the stochastic distribution of
returns using the information provided on the LC’s web site. Although we cannot observe the
investors’ beliefs, we test this assumption indirectly by exploiting a feature of the LC environment:
Many investors choose LC loans both manually and through an optimization tool based on returns
and default probabilities posted on the LC site. We use the sample of investors that use both
methods to show that our estimate of investor-specific risk aversion obtained from the manually
chosen component of their LC portfolio does not differ significantly from the estimate obtained from
the component of the portfolio chosen with the tool. Testing the validity of assumptions concerning
person-specific priors on idiosyncratic risk is typically impossible in real-life environments. Being
able to validate our assumptions, although in an admittedly indirect way, represents an important
advantage of this setting.
With this method we estimate parameters of risk aversion for each investor and portfolio choice
in our sample. The average ARA implied by the tradeoff between expected return and idiosyncratic
risk in our sample of portfolio choices is 0.037. Our estimates imply an average income-based
Relative Risk Aversion (income-based RRA) of 2.81, with substantial unexplained heterogeneity
and skewness.7
In estimating risk attitudes based on idiosyncratic risk, our approach is similar to that of
the economic literature that studies risk attitudes and insurance choices (Cohen and Einav, 2007;
Barseghyan et al., 2013; Einav et al., 2012). Our estimates are also comparable to experimental
measures of risk aversion, as investors in our model face choices similar to those faced by experimen-
tal subjects along important dimensions. Our model transforms a complex portfolio choice problem
into a choice between well-defined lotteries of pure idiosyncratic risk, where returns are character-
ized by a discrete failure probability (i.e., default), and the stakes are small relative to total wealth
(the median investment in LC is $375).8 The level, distribution, and skewness of the estimated
7The income-based RRA is a risk preference parameter often reported in the experimental literature. It is obtainedassuming that the investor’s outside wealth is zero; that is: ARA ·E[y], where E[y] is the expected income from thelottery offered in the experiment.
8For estimation of risk aversion in real life environments with idiosyncratic risk, see also Jullien and Salanie (2000),
4
risk aversion parameters are similar to those obtained in laboratory and field experiments.9 These
similarities indicate that investors in our sample, despite being a self-selected sample of individuals
who invest on-line, have risk preferences similar to individuals in other settings.
We show that our method generates consistent estimates for the curvature of the utility function
under the expected utility framework and alternative preference specifications, such as loss aversion
and narrow framing, as in Barberis and Huang (2001) and Barberis et al. (2006). This is a key
feature of our estimation procedure since under the expected utility framework the observed high
levels of risk aversion in small stake environments are difficult to reconcile with the observable be-
havior of agents in environments with larger stakes (Rabin, 2000). We also show that our estimates
of the elasticity of ARA with respect to investor’s wealth characterize agents’ risk attitudes, both,
in large and small stake environments.
We exploit the panel dimension of our data to derive the cross-sectional and within-investor
elasticities of risk aversion with respect to wealth. First, in the cross section, we find that wealthier
investors exhibit lower ARA and higher RRA. Since we use imputed net worth as a proxy for wealth,
our preferred specification corrects for measurement error in the wealth proxy using house prices
in the investor’s zip code as an instrument. We obtain an elasticity of ARA to wealth of -0.074,
which implies a cross sectional wealth elasticity of the RRA of 0.93.10 These findings coincide
qualitatively with Guiso and Paiella (2008) and Cicchetti and Dubin (1994), who also base their
analysis on risk aversion parameters estimated independently of their wealth measure.
Second, we estimate the elasticity of risk aversion to changes in wealth in an investor fixed-effect
specification to characterize the shape of the utility function for the average investor. We use the
decline in house prices in the investor’s zip code during our sample period—October 2007 to April
2008—as a source of variation in investor wealth. The results indicate that the average investor’s
RRA increases after experiencing a negative housing wealth shock, with an estimated elasticity of
−1.85. This is consistent with investors exhibiting decreasing Relative Risk Aversion, and with
theories of habit formation (as in Campbell and Cochrane, 1999) and incomplete markets (as in
Jullien and Salanie (2008), Bombardini and Trebbi (2012), Harrison, Lau and Towe (2007), Chiappori et al. (2008),Post et al. (2008), Chiappori et al. (2009), and Barseghyan et al. (2011).
9See for example Barsky et al. (1997), Holt and Laury (2002), Choi et al. (2007), and Harrison, Lau and Rutstrom(2007).
10The wealth-based RRA is not directly observable, we compute its elasticity from the following relationship:ξRRA,W = ξARA,W + 1,where ξRRA,W and ξARA,W refer to the wealth elasticities of RRA and ARA, respectively.
5
Guvenen, 2009).11
The contrasting signs of the cross-sectional and within-investor elasticities confirm that investors
have heterogeneous risk preferences that are correlated with wealth in the cross section. This implies
that inference on the elasticity of risk aversion to wealth from cross sectional data will be biased.
In our setting, the bias implied by the joint distribution of preferences and wealth is large enough
to flip the sign of the investor-specific risk aversion elasticity to wealth.12
A novel feature of our estimation approach is that it disentangles the measurement of risk
aversion from investors’ assessments about the systematic risk of LC loans. On average, investors’
perceived systematic risk premium in LC increases from 5.7% to 8.9% between the first and last
months of the sample period. The increase in the perceived systematic risk is concurrent with
the drop in house prices, which indicates that wealth shocks are potentially correlated with general
changes in investors’ beliefs. This calls for caution when using the share of risky assets as a measure
of investor RRA. In our context, inference based on the share of risky assets alone would have led
us to overestimate the investor-specific elasticity of risk aversion to wealth.
The rest of the paper is organized as follows. Section 2 describes the Lending Club platform.
Section 3 illustrates the portfolio choice model and sets out our estimation strategy. Section 4
describes the data and the sample restrictions. Section 5 presents and discusses the empirical
results and provides a test of the identification assumptions. Section 6 explores the relationship
between risk preferences and wealth. And Section 7 concludes.
2 The Lending Platform
Lending Club (LC) is an online U.S. lending platform that allows individuals to invest in portfolios
of small loans. The platform started operating in June 2007. As of May 2010, it has funded
$112,003,250 in loans and provided an average net annualized return of 9.64% to investors.13 Below,
11These preference specifications are also consistent with the empirical findings in Calvet et al. (2009). Brunnermeierand Nagel (2008), on the other hand, find support for CRRA.
12Chiappori and Paiella (2011) find the bias from the cross sectional estimation to be economically insignificantwhen risk aversion is measured using the share of risky assets. Tanaka et al. (2010) use rainfall across villages inVietnam as an instrument for wealth in the cross section and find significant difference between the OLS and IVestimators. However, to obtain the elasticity of the agent-specific risk aversion, they must assume that preferencesare equal across villages otherwise.
13The median investor in our sample period is a retail investor. The composition then shifted towards institutionalinvestors, a common trend in the peer-to-peer industry. As of 2014, 80% of investment going into P2P platformsProsper and Lending Club is from institutional investors (Financial Times, October 5, 2014). For the latest figures
6
we provide an overview of the platform and derive the expected return and variance of investors’
portfolio choices.
2.1 Overview
Borrowers need a U.S. SSN and a FICO score of 640 or higher in order to apply. They can request
a sum ranging from $1,000 to $25,000, usually to consolidate credit card debt, finance a small
business, or fund educational expenses, home improvements, or the purchase of a car.
Each application is classified into one of 35 risk buckets based on the FICO score, the requested
loan amount, the number of recent credit inquiries, the length of the credit history, the total and
currently open credit accounts, and the revolving credit utilization, according to a pre-specified
published rule posted on the website.14 LC also posts a default rate for each risk bucket, taken
from a long term validation study by TransUnion, based on U.S. unsecured consumer loans. All the
loans classified in a given bucket offer the same interest rate, assigned by LC based on an internal
rule.
A loan application is posted on the website for a maximum of 14 days. It becomes a loan only
if it attracts enough investors and gets fully funded. All the loans have a 3 year term with fixed
interest rates and equal monthly installments, and can be prepaid with no penalty for the borrower.
When the loan is granted, the borrower pays a one-time fee to LC ranging from 1.25% to 3.75%,
depending on the risk bucket. When a loan repayment is more than 15 days late, the borrower is
charged a late fee that is passed to investors. Loans with repayments more than 120 days late are
considered in default, and LC begins the collection procedure. If collection is successful, investors
receive the amount repaid minus a collection fee that varies depending on the age of the loan and
the circumstances of the collection. Borrower descriptive statistics are shown in Table 1, panel A.
Investors in LC allocate funds to open loan applications. The minimum investment in a loan is
$25. According to a survey of 1,103 LC investors in March 2009, diversification and high returns
relative to alternative investment opportunities are the main motivations for investing in LC.15 LC
refer to: https://www.lendingclub.com/info/statistics.action.14Please refer to https://www.lendingclub.com/info/how-we-set-interest-rates.action for the details of the classifi-
cation rule and for an example.15To the question “What would you say was the main reason why you joined Lending Club”, 20% of respondents
replied “to diversify my investments”, 54% replied “to earn a better return than (...)”, 16% replied “to learn moreabout peer lending”, and 5% replied “to help others”. In addition, 62% of respondents also chose diversification andhigher returns as their secondary reason for joining Lending Club.
7
lowers the cost of investment diversification inside LC by providing an optimization tool that con-
structs the set of efficient loan portfolios for the investor’s overall amount invested in LC—i.e., the
minimum idiosyncratic variance for each level of expected return (see Figure 1).16 In other words,
the tool helps investors to process the information on interest rates and default probabilities posted
on the website into measures of expected return and idiosyncratic variance, that may otherwise
be difficult to compute for an average investor (these computations are performed in Subsection
2.2).17 When investors use the tool, they select, among all the efficient portfolios, the preferred
one according to their own risk preferences. Investors can also use the tool’s recommendation as
a starting point and then make changes. Or they can simply select the loans in their portfolio
manually.
Of all portfolio allocations between LC’s inception and June 2009, 39.6% was suggested by the
optimization tool, 47.1% was initially suggested by the tool and then altered by the investor, and
the remaining 13.3% was chosen manually.18
Given two loans that belong to the same risk bucket (with the same idiosyncratic risk), the
optimization tool suggests the one with the highest fraction of the requested amount that is already
funded. This tie-breaking rule maximizes the likelihood that loans chosen by investors are fully
funded. In addition, if a loan is partially funded at the time the application expires, LC provides
the remaining funds.
2.2 Return and Variance of the Risk Buckets
This subsection derives the expected return and variance of individual loans and risk buckets,
following the same assumptions as the LC platform. All the loans in a given risk bucket z = 1, ..., 35
are characterized by the same scheduled monthly payment per dollar borrowed, Pz, over the 3 years
(36 monthly installments). The per dollar scheduled payment Pz and the bucket specific default
rate πz fully characterize the expected return and variance of per project investments, µz and σ2z .
LC considers a geometric distribution for the idiosyncratic monthly survival probability of the
16During the period analyzed in this paper, the portfolio tool appeared as the first page to the investors. LC hasrecently changed its interface and, before the portfolio tool page, it has added a stage where the lender can simplypick between 3 representative portfolios of different risk and return.
17The tool normalizes the idiosyncratic variance into a 1–0 scale. Thus, while the tool provides an intuitive sortingof efficient portfolios in terms of their idiosyncratic risk, investors always need to analyze the recommended portfoliosof loans to understand the actual risk level imbedded in the suggestion.
18We exploit this variation in Subsection 5.2 to validate the identification assumptions.
8
individual projects: The probability that the loan survives until month τ ∈ [1, 36] is Pr (T = τ) =
πz (1− πz)τ .19 The resulting expectation and variance of the present value of the payments, Pz, of
any project in bucket z are:
µz = Pz
[1−
(1− πz1 + r
)36]
1− πzr + πz
σ2z =
35∑t=1
πz (1− πz)t(
t∑τ=1
Pz(1 + r)τ
)2
+
(36∑τ=1
Pz(1 + r)τ
)2
(1− πz)36 − µ2z
where r is the risk-free interest rate.
The idiosyncratic risk associated with bucket z decreases with the level of diversification within
the bucket; that is, the number of projects from bucket z in the portfolio of investor i, niz. The
resulting idiosyncratic variance is therefore investor specific:
var[riz]
=σ2zniz. (1)
where riz is the idiosyncratic component of the bucket’s return, Riz
Then, the variance of the return on investment in bucket z can be decomposed as follows:
var[Riz]
= V iz +
σ2zniz
where V iz corresponds to the bucket’s non-diversifiable risk and σ2z/n
iz is its idiosyncratic component.
The expected return of an investment in bucket z is not affected by the number of loans in the
investor’s portfolio; it is equal to the expected return on a loan in bucket z, µz, and is constant
across investors:20
E [Rz] = E[Riz]
= µz. (2)
19Although LC does not explicitly state in the web page that the probabilities of default are idiosyncratic, theoptimization tool used to calculate the set of minimum variance portfolios works under this assumption.
20The analysis in subsection 5.2 confirms that investors’ beliefs about the probabilities of default do not differsubstantially from those posted on the website and, therefore, σz and µz are constant across investors.
9
3 Estimation Procedure
To capture the fact that most individuals invest not only in the market and the risk free asset but
also in individual securities, our framework considers investors that, instead of simply holding a
replica of the market portfolio, also hold securities based on their own subjective insights that they
will generate an excess return. The decision to invest in LC depends on investors’ knowledge of
its existence and their subjective expectation that LC is, indeed, a good investment opportunity.
Thus, it is reasonable to assume that investors in LC have special insights, which explains why, as
we show later, their portfolio departs from just replicating the market.21
Our framework starts by recognizing that there is a high degree of comovement between se-
curities, and specifically to our case, the probability of default of the loans in LC is potentially
correlated with macroeconomic fluctuations. We use the Sharpe Diagonal Model to capture this
feature and assume that returns are related only through a common systematic factor (i.e., market
or macroeconomic fluctuations). Under this assumption, returns on LC loans can be decomposed
into a common systematic factor and a component orthogonal to the market (we also refer to it as
independent component).
The advantage of this model is that the optimal portfolio of LC loans depends only on the
expected return and variance of the independent component, as the next subsection shows. In other
words, the optimal amount invested in each LC bucket does not depend on the return covariance
with the investor’s overall risk exposure, nor does it require knowing the amount and characteristics
of her outside wealth.
We test the assumptions on investors’ beliefs in Section 5, i.e. that they act as if there is a
common systematic factor, and that the choice of LC loans does not simply replicate the market
portfolio. The data strongly support both hypotheses.
3.1 The Model
Each investor i chooses the share of wealth to be invested in z = 1, ..., Z securities with return Rz
and Z ≥ 35. We consider securities z = 1, ..., 35 to represent the LC risk buckets. And we explicitly
include outside investment options, z = 36, ..., Z, that represent the market composite, a risk-free
asset, real estate, human capital, and other securities potentially unknown to the econometrician.
21This framework is equivalent to the one analyzed in Treynor and Black (1973).
10
A projection of the return of each security against that of the market, Rm, gives two factors.
The first is the systematic return of the security, and the second its independent return:
Rin = βin ·Rm + rin. (3)
We assume all LC risk buckets have the same systematic risk, and allow the prior about such
systematic component, βiL, to be investor specific. That is, for all z = 1, ..., 35 : βiz = βiL. We test
and validate empirically this assumption in Subsection 5.2.22
The assumption that all risk buckets have the same systematic risk is combined with the Sharpe
Diagonal Model, which assumes that the returns on the LC loans are related to other securities and
investment opportunities only through their relationships with a common underlying factor, and
thus the independent returns, defined in equation (3), are uncorrelated. Allowing a time dimension,
the independent returns are also uncorrelated across time. This feature of the model is exploited
in Subsection 6.2, when we analyze multiple investment by the same agent.
Assumption 1. Sharpe Diagonal Model.
for all n 6= h : cov[rin, r
ih
]= 0
An example of an underlying common factor is an increase in macroeconomic risk (i.e., financial
crisis) triggering correlated defaults across buckets. In our setting, such a common factor is reflected
in the systematic component of equation (3) and can vary across investors and over time.
Each investor i faces the following portfolio choice problem
max{xz}Zz=1
Eu
(W i
[Z∑z=1
xizRiz
])
s.t.Z∑z=1
xz = 1
xz = 0 or xz ≥ 25 for all z = 1, ..., 35.
{xz}Zz=1 correspond to the shares of the wealth, Wi, invested in each security z.
22Note that under this assumption, the prior about the systematic risk Vz introduced in Subsection 2.2 is investorspecific and it is given by V i
z =(βiL
)2 · var [Rm], for all z = 1, ..., 35.
11
The first order condition characterizing the optimal portfolio share of any LC bucket z = 1, ..., 35
is:
foc (xz) : E[u′ (ci) ·W i ·Rz
]− µi − λih (xz > 25) = 0
where µi corresponds to the multiplier on the budget constraint,∑Z
z xz = 1, and λiz is the Khun-
Tucker multiplier on the minimum LC investment constraint, which is zero for all those buckets for
which the investor has a positive position.
For all buckets with xz ≥ 25, a first-order linearization on the first order condition around
expected consumption results in the following optimality condition for all buckets with positive
investment:
foc(xz) : u′(E[ci]
)W iE
[Riz]
+ u′′(E[ci])(W i)2
(Z∑
z′=1
xiz′cov[Riz′ , R
iz
])− µi = 0 (4)
Due to the $25 minimum investment per loan constraint, this first order condition characterizes
the optimal portfolio shares in those risk buckets where there is an interior solution, e.g. where the
investor chooses a positive and finite investment amount. In other words, this optimality condition
describes the investment share that makes the investor indifferent about allocating an additional
dollar to a loan in bucket z when investment in the bucket is greater than zero.23 In contrast, in
corner solutions where the additional expected return is always smaller (larger) than the marginal
increase in risk for any investment amount larger than $25, investment in that bucket will be zero
(infinity). We describe in detail the optimality condition that characterizes interior and corner
solutions in the on-line appendix B.
Under assumption 1, for a given investor i, the covariances between any two LC bucket returns
and between the returns on any LC bucket and outside security are constant across risk buckets.
In particular, given that their comovement is given by a common macroeconomic factor (i.e., the
23Since our estimation procedure exploits only buckets with positive investments, we use this margin of choice as anindependent test of preference consistency. We show in the on-line appendix C.2 that the estimated risk preferencesare consistent with those implied by the forgone buckets.
12
market return), they can be expressed as follows:
∀z ∈ {1, ..., 35} ∧ ∀h ∈ {36, ..., Z} : cov[Riz, Rih] = βiLβ
ihvar[Rm]
∀z, z′ ∈ {1, ..., 35}, z 6= z′ : cov[Riz, Riz′ ] = (βiL)2var[Rm]
∀z ∈ {1, ..., 35} : cov[Riz, Riz] = (βiL)2var[Rm] + var[riz]
where βiL is the market sensitivity, or beta, of the LC returns, defined in equation (3) and assumed
constant across buckets; and βih is the corresponding sensitivity for the investor’s outside security
h. The variance of any risk bucket z is given by its systematic component, V iz = (βiL)2var[Rm],
and its idiosyncratic risk, derived in equation (1): var[riz] = σ2z/niz.
Rearranging terms, we derive our main empirical equation. For all LC risk buckets z = 1, ..., 35
for which investor i has a positive position:
E[Rz] = θi +ARAi · Wixizniz
· σ2z (5)
where:
θi ≡ µi
W iu′(E[ci])+ARAi W i βiL var[Rm]
(βiL
35∑z′=1
xiz′ +
Z∑z′=36
βiz′xiz′
)(6)
ARAi ≡ −u′′(E[ci])
u′ (E [ci])(7)
The parameter θi reflects the investor’s assessment of the systematic component of the LC invest-
ment.24 It is investor and investment-specific, and constant across buckets for each investment. The
parameter ARAi corresponds to the Absolute Risk Aversion. It captures the extra expected return
needed to leave the investor indifferent when taking extra risk. Such parameter refers to the local
curvature of the preference function and it is not specific to the expected utility framework. We
show in on-line appendix A that the same equation characterizes the optimal LC portfolio and the
curvature of the utility function under two alternative preference specifications: 1) when investors
are averse to losses in their overall wealth, and 2) when investor utility depends in a non-separable
24If one of the outside securities in the choice set of the investor is a risk-free asset with return Rf , then θi ≡Rf +ARAiW iβi
Lvar[Rm](βiL
∑35z′=1 x
iz′ +
∑Zz′=36 β
iz′x
iz′
).
13
way on both the overall wealth level and the income flow from specific components of the portfolio
(narrow framing).
The optimal LC portfolio depends only on the investor’s risk aversion, and the expectation and
variance of the independent return of each bucket z. The holding of a portfolio of securities outside
LC, which includes market systematic fluctuation, optimally adjusts to account for the indirect
market risk embedded in LC.25
Finally, we cannot compute the Relative Risk Aversion (RRA) without observing the expected
lifetime wealth of the investors.26 However, for the purpose of comparing our estimates with the
results from other empirical studies on relative risk aversion, we follow that literature and define
the relative risk aversion parameter based solely on the income generated by investing in LC (see,
for example, Holt and Laury, 2002). This income-based RRA is defined as follows:
ρi ≡ ARAi · IiL ·(E[RiL]− 1)
(8)
where IiL is the total investment in LC, IiL = W i∑35
z=1 xiz, and E
[RiL]
is the expected return on
the LC portfolio, E[RiL]
=∑35
z=1 xizE [Rz].
4 Data and Sample
Our sample covers the period between October 2007 and April 2008. Below we provide summary
statistics of the investors’ characteristics and their portfolio choices, and a description of the sample
construction.
4.1 Investors
For each investor we observe the home address zip code, verified by LC against the checking account
information, and age, gender, marital status, home ownership status, and net worth, obtained
through Acxiom, a third party specialized in recovering consumer demographics. Acxiom uses a
proprietary algorithm to recover gender from the investor names, and matches investor names and
25See Treynor and Black (1973) for the original derivation of this result.26Although we cannot compute RRA, in Section 6 we show that we can infer its elasticity with respect to wealth,
based on the elasticity of ARA: ξRRA,W = ξARA,W + 1.
14
home addresses to available public records to recover age, marital status, home ownership status,
and an estimate of net worth. Such information is available at the beginning of the sample.
Table 1, panel B, shows the demographic characteristics of the LC investors. The average in-
vestor in our sample is 43 years old, 8 years younger than the average respondent in the Survey of
Consumer Finances (SCF). As expected from younger investors, the proportion of married partic-
ipants in LC (56%) is lower than in the SCF (68%). Men are over-represented among participants
in financial markets, they account for 83% of the LC investors; similarly, the faction of male respon-
dents in the SCF is 79%. In terms of income and net worth, investors in LC are comparable to other
participants in financial markets, who are typically wealthier than the median U.S. households. The
median net worth of LC investors is estimated between $250,000 and $499,999, significantly higher
then the median U.S. household ($120,000 according to the SCF), but similar to the estimated
wealth of other samples of financial investors. Korniotis and Kumar (2011), for example, estimate
the wealth of clients in a major U.S. discount brokerage house in 1996 at $270,000.
To obtain an indicator of housing wealth, we match investors’ information with the Zillow
Home Value Index by zip code. The Zillow Index for a given geographical area is the value of
the median property in that location, estimated using a proprietary hedonic model based on house
transactions and house characteristics data, and it is available at a monthly frequency. LC investors
are geographically disperse but tend to concentrate in urban areas and major cities. Table 1 shows
the descriptive statistics of median house values on October 2007 and their variation during the
sample period—October 2007 to April 2008.
4.2 Sample Construction
We consider as a single portfolio choice all the investments an individual makes within a calendar
month.27 The full sample contains 2,168 investors, 5,191 portfolio choices, which results in 50,254
investment-bucket observations. To compute the expected return and idiosyncratic variance of the
investment-bucket in equations (1) and (2), we use as the risk free interest rate, the 3-year yield on
Treasury Bonds at the time of the investment. Table 2, panel A, reports the descriptive statistics
of the investment-buckets. The median expected return is 12.2%, with an idiosyncratic variance of
27This time window is arbitrary and modifying it does not change the risk aversion estimates. We chose a calendarmonth for convenience, since it coincides with the frequency of the real estate price data that we use to proxy forwealth shocks in the empirical analysis.
15
3.5%. Panel B, describes the risk and return of the investors’ LC portfolios. The median portfolio
expected return in the sample is 12.1%, almost identical to the expectation at the bucket level, but
the idiosyncratic variance is substantially lower, 0.005%, due to risk diversification across buckets.
Our estimation method imposes two requirements for inclusion in the sample. First, estimating
risk aversion implies recovering two investor specific parameters from equation (5). Therefore, a
point estimate of the risk aversion parameter can only be recovered when a portfolio choice contains
more than one risk bucket.
Second, our identification method relies on the assumption that all projects in a risk bucket
have the same expected return and variance. Under this assumption investors will always prefer
to exhaust the diversification opportunities within a bucket, i.e., will prefer to invest $25 in two
different loans belonging to bucket z instead of investing $50 in a single loan in the same bucket.
It is possible that some investors choose to forego diversification opportunities if they believe that
a particular loan has a higher return or lower variance than the average loan in the same bucket.
Because investors’ private insights are unobservable to the econometrician, such deviations from full
diversification will bias the risk aversion estimates downwards. To avoid such bias we exclude all
non-diversified components of an investment. Thus, the sample we base our analysis on includes:
1) investment components that are chosen through the optimization tool, which automatically
exhausts diversification opportunities, and 2) diversified investment components that allocate no
more than $50 to any given loan.
After imposing these restrictions, the analysis sample has 2,168 investors and 3,745 portfolio
choices. The descriptive statistics of the analysis sample are shown in Table 2, column 2. As
expected, the average portfolio in the analysis sample is smaller and distributed across a larger
number of buckets than the average portfolio in the full sample. The average portfolio expected
return is the same across the two samples, while the idiosyncratic variance in the analysis sample is
smaller. This is expected since the analysis sample excludes non-diversified investment components.
In the wealth analysis, we further restrict the sample to those investors that are located in zip
codes where the Zillow Index is computed. This reduces the sample to 2,061 investors and 3,405
portfolio choices. This final selection does not alter the observed characteristics of the portfolios sig-
nificantly (Table 2, column 3). To maintain a consistent analysis sample throughout the discussion
that follows, we perform all estimations using this final subsample unless otherwise noted.
16
5 Risk Aversion Estimates
Our baseline estimation specification is based on equation (5). We allow for an additive error term,
such that for each investor i we estimate the following equation:
E [Rz] = θi +ARAi · Wixizniz
· σ2z + εiz (9)
There is one independent equation for each bucket z with a positive investment in the investor’s
portfolio. The median portfolio choice in our sample allocates funding to 10 buckets, which provides
us with multiple degrees of freedom for estimation. We estimate the parameters of equation (9)
with Ordinary Least Squares.
Panels (a) and (b) in Figure 2 show two examples of portfolio choices. The vertical axis measures
the expected return of a risk bucket, E [Rz], and the horizontal axis measures the bucket variance
weighted by the investment amount, W ixizσ2z/n
iz. The slope of the linear fit is our estimate of the
absolute risk aversion and it is reported on the top of each plot.
The error term captures deviations from the efficient portfolio due to measurement errors by
investors, and real or perceived private information. The OLS estimates will be unbiased as long
as the error component does not vary systematically with bucket risk. We discuss and provide
evidence in support of this identification assumption below.
5.1 Results
The descriptive statistics of the estimated parameters of equation (9) for each portfolio choice are
presented in Table 3. The average estimated ARA across all portfolio choices is 0.037 (column 1).
Investors exhibit substantial heterogeneity in risk aversion, and its distribution is left skewed: the
median ARA is 0.044 and the standard deviation 0.0245. This standard deviation overestimates the
standard deviation of the true ARA parameter across investments because it includes the estimation
error that results from having a limited number of buckets per portfolio choice. Following Arellano
and Bonhomme (2012), we can recover the variance of the true ARA by subtracting the expected
estimation variance across all portfolio choices. The calculated standard deviation of the true ARA
is 0.0226, indicating that the estimation variance is small relative to the variance of risk aversion
17
across investments.28 The range of the ARA estimates is consistent with the estimates recovered
in the laboratory. Holt and Laury (2002), for example, obtain ARA estimates between 0.003 and
0.109, depending on the size of the bet.
The experimental literature often reports the income-based RRA, defined in equation (8). To
compare our results with those of laboratory participants, we report the distribution of the implied
income-based RRA in Table 3, column 4. The mean income-based RRA is 2.81 and its distribution
is right-skewed (median 1.62). This parameter scales the measure of absolute risk aversion according
to the lottery expected income; therefore, it mechanically increases with the size of the bet. Column
3 of Table 3 reports the distribution of expected income from LC. The mean expected income is
$125.6, substantially higher than the bet in most laboratory experiments. Not surprisingly, although
the computed ARA in experimental work is typically larger than our estimates, the income-based
RRA parameter is smaller, ranging from 0.3 to 0.52 (see for example Chen and Plott, 1998, Goeree
et al., 2002, Goeree et al., 2003, and Goeree and Holt, 2004). Our results are comparable to Holt
and Laury (2002), who also estimate risk aversion for agents facing large bets and (implicitly) find
income-based RRA similar to ours, 1.2. Finally, Choi et al. (2007) report risk premia with a mean
of 0.9, which corresponds to an income-based RRA of 1.8 in our setting. That paper also finds
right skewness in their measure of risk premia.
Our findings imply that the high levels of risk aversion exhibited by subjects in laboratory
experiments extrapolate to actual small-stake investment choices. Rabin and Thaler (2001, 2002)
emphasize that such levels of risk aversion with small stakes are difficult to reconcile, within the
expected utility framework over total wealth, with the observable behavior of agents in environments
with larger stakes. Our results are subject to that critique. The median RRA computed using with
our estimates of ARA and investors’ net-worth is 8,858 (average is 26,317).29 This suggests that
EU framework on overall wealth cannot describe agents behavior in our environment. We show
in on-line appendix A that the ARA estimated here describes the curvature of the utility function
in other preference frameworks that are consistent with observed risk behavior over small and
28The variance of the true ARA is calculated as:
var[ARAi
]= var
[ARA
i]− E
[σ2ARAi
]where the first term is the variance of the OLS ARA point estimates across all investments, and the second term isthe average of the variance of the OLS ARA estimates across all investments.
29The Net-Worth based RRA reported here is computed as RRAi = ARAi ·NWi, where NWi is the mean pointof the net worth interval that Acxiom assigns to each investor in our sample.
18
large stake gambles. In particular, on-line appendix A.1 describes the optimal portfolio choice in
a behavioral model in which utility depends (in a non-separable way) on both overall wealth level,
W , and the flow of income from specific components of the agent’s portfolio, y.30 In such alternate
preference specifications, agents’ ARA varies with, both, the level of overall wealth and the income
flow generated by the gamble. This implies that the estimated level of ARA may be larger for
small-stake gambles (i.e., ∂ARA(y,W )/∂y < 0 ). Nevertheless, the elasticity of ARA with respect
to investor’s wealth (i.e., ∂ARA(y,W )∂W
WARA(y,W )), our focus in the next section, is consistent in small
and large stake environments.
Column 2 in Table 3 shows the estimates of the parameter θ, defined in equation (6), which
captures the systematic component of LC. In our framework, the systematic component is driven
by the common covariance between all LC bucket returns and the market, βL, or any potential
risk of the lending platform common to all risk buckets. The average estimated θ is 1.086, which
indicates that the average investor requires a systematic risk premium of 8.6%. The estimated θ
presents very little variation in the cross section of investors (coefficient of variation 3%), when
compared to the variation in the ARA estimates (coefficient of variation of 66% ).31 Note that
our ARA estimates are not based on this systematic risk premium; instead, they are based on the
marginal premium required to take an infinitesimally greater idiosyncratic risk.
Table 4 presents the average and standard deviation of the estimated parameters by month.
The average ARA increases from 0.029 during the first three months, to 0.039 during the last three
(column 1). This average time series variation is potentially due to heterogeneity across investors
as well as within investor variation, since not all investors participate in LC every month. The
analysis in the next section disentangles the two sources of variation.
The estimated θs, shown in column 2, imply that the average systematic risk premium increases
from 5.7% to 8.9% between the first and last three months of the sample period. Note that
the LC web page provides no information on the systematic risk of LC investments. Thus, this
change is solely driven by changes in investors’ beliefs about the potential systematic risk of the
30This is in line with Barberis and Huang (2001) and Barberis et al. (2006), which propose a framework whereagents exhibit loss aversion over changes in specific components of their overall portfolio, together with decreasingrelative risk aversion over their entire wealth. In the expected utility framework, Cox and Sadiraj (2006) propose autility function with two arguments (income and wealth) where risk aversion is defined over income, but it is sensitiveto the overall wealth level.
31As with the ARA, the estimation variance is small relative to the variance across investments. The standarddeviation of θ is 0.028, while the standard deviation of θ after subtracting the estimation variance is 0.0271.
19
lending platform; that is, correlation between the likelihood of default of LC loans and aggregate
macroeconomic shocks (covariance between LC returns and market returns, βL), or about the
expected market risk premium (E [Rm]), or about the functioning of the platform. This pattern
indicates that wealth shocks are potentially correlated with changes in investors’ beliefs about risk
and return on financial assets. Thus, we cannot infer the elasticity of RRA to wealth by observing
changes in the share of risky assets after a wealth shock, as they may be simply reflecting changes
in beliefs about the underlying distribution of risky returns. Our proposed empirical strategy in
the next section overcomes this identification problem.
5.2 Belief Heterogeneity and Bias: The Optimization Tool
Above we interpret the observed heterogeneity of investor portfolio choices as arising from differ-
ences in risk preferences. Such heterogeneity may also arise if investors have different beliefs about
the risk and returns of the LC risk buckets. Note that differences in beliefs about the systematic
component of returns will not induce heterogeneity in our estimates of the ARA. This type of belief
heterogeneity will be captured by variations in θ across investors.
The parameter θ will not capture heterogeneity of beliefs that affects the relative risk and
expected return across buckets. This is the case if investors believe the market sensitivity of
returns to be different across LC buckets, i.e. if βiz 6= βiL for some z = 1, ..., 35; or if investors’ priors
about the stochastic properties of the buckets idiosyncratic return differ from the ones computed
in equations (1) and (2), i.e. Ei [Rz] 6= E [Rz] or σiz 6= σz for some z = 1, ..., 35. In such cases, the
equation characterizing the investor’s optimal portfolio is given by:
E [Rz] = θi +[ARAi ·Bi
σ +Biµ +Bi
β
]· W
ixizniz
σ2z
This expression differs from our main specification equation (5) in three bias terms: Bσ ≡(σiz/σz
)2,
Bµ ≡(E [Rz]− Ei [Rz]
)/(W ixizσ
2z/n
iz
), and Bβ ≡
(βiz − βiL
)/(W ixizσ
2z/n
iz).
Two features of the LC environment allow us to estimate the magnitude of the overall bias from
these sources. First, LC posts on its website an estimate of the idiosyncratic default probabilities
for each bucket. Second, LC offers an optimization tool to help investors diversify their loan
portfolio. The tool constructs the set of efficient loan portfolios, given the investor’s total amount
20
in LC—i.e., the minimum idiosyncratic variance for each level of expected return. Investors then
select, among all the efficient portfolios, the preferred one according to their own risk preferences.
Importantly, the tool uses the same modeling assumptions regarding investors’ beliefs that we use
in our framework: the idiosyncratic probabilities of default are the ones posted on the website and
the systematic risk is common across buckets, i.e. βz = βL.32
Thus, we can measure the estimation bias by comparing, for the same investment, the ARA
estimates obtained independently from two different components of the portfolio choice: the loans
suggested by the tool and those chosen manually. If investors’ beliefs do not deviate systematically
across buckets from the information posted on LC’s website and from the assumptions of the
optimization tool, we should find investor preferences to be consistent across the two measures.
Note that our identification assumption does not require that investors agree with LC assumptions.
It suffices that the difference in beliefs does not vary systematically across buckets. For example,
our estimates are unbiased if investors believe that the idiosyncratic risk is 20% higher than the
one implied by the probabilities reported in LC, across all buckets. Note, moreover, that our test
is based on investors’ beliefs at the time of making the portfolio choices. These beliefs need not to
be correct ex post.
For this test we use the subsample of investments that combine buckets chosen through the
optimization tool with buckets chosen manually. Then, for each investment, we independently
compute the risk aversion implied by the component suggested by the optimization tool (Automatic
buckets) and the risk aversion implied by the component chosen directly by the investor (Non-
Automatic buckets). Panels (c) and (d) in Figure 2 provide an example of this estimation. Both
panels plot the expected return and weighted idiosyncratic variance for the same portfolio choice.
Panel (c) includes only the Automatic buckets, suggested by the optimization tool. Panel (d)
includes only the Non Automatic buckets, chosen directly by the investor. The estimated ARA
using the Automatic and Non-Automatic bucket subsamples are 0.048 and 0.051 respectively for
this example.
We perform the independent estimation above for all portfolio choices that have at least two
Automatic and two Non-Automatic buckets. To verify that investments that contain an Automatic
component are representative of the entire sample, we compare the extreme cases where the entire
32See on-line appendix B.1 for the derivation of the efficient portfolios suggested by the optimization tool.
21
portfolio is suggested by the tool and those where the entire portfolio is chosen manually. The
median ARA is 0.0455 and 0.0446 respectively, and the mean difference across the two groups is
not statistically significant at the standard levels. This suggests that our focus in this subsection
on investments with Automatic and Non-Automatic components is representative of the entire
investment sample.
Table 5, panel A, reports the descriptive statistics of the ARA estimated using the Automatic
and the Non-Automatic buckets. The average ARA is virtually identical across the two estimations
(Table 5, columns 1 and 2). We calculate, for each investment, the difference between the two ARA
estimates; Figure 3a shows the entire distribution of this difference and Column 3 of Table 5 reports
its descriptive statistics. The mean is zero and the distribution of the difference is concentrated
around zero, with kurtosis 11.53. This implies that the bias is close to zero not only in expectation,
but investment-by-investment.
These results suggest that investors’ beliefs about the stochastic properties of the loans in LC do
not differ substantially from those posted on the website. They also suggest that investors’ choices
are consistent with the assumption that the systematic component is constant across buckets.
Overall, these findings validate the interpretation that the observed heterogeneity across investor
portfolio decisions is driven by differences in risk preferences.
In Table 5, panels B and C, we show that the difference in the distribution of the estimated
ARA from the automatic and non-automatic buckets is insignificant both during the first and second
halves of the sample period. This is key for interpreting the results in the next section, where we
explore how the risk aversion estimates change in the time series with changes in house prices.
Although we cannot directly rule out that changes in house prices are correlated with changes
in beliefs for the same investor, we find that beliefs remain consistent with our assumptions in
expectation throughout the sample period.
Table 5, columns 4 through 6, show that the estimated risk premia, θ, also exhibit almost
identical mean and standard deviations when obtained independently using the Automatic and
Non-Automatic investment components. Column 6 and Figure 3b show the distribution of the
difference between the two estimates of θ for the same investment. They suggest that our estimates
of the risk premium are unbiased.
It is worth reiterating that these findings do not imply that investors’ beliefs about the overall
22
risk of investing in LC do not change during the sample period. On the contrary, the observed
average increase in the estimated systematic risk premium in Table 4 is also observed in panels B
and C of Table 5: θ increases by 2.5 percentage points between the first and second halves of the
sample. The results in Table 5 imply that changes in investors’ beliefs are fully accounted for by
a common systematic component across all risk buckets and, thus, do not bias our risk aversion
estimates.
6 Risk Aversion and Wealth
This section explores the relationship between investors’ risk taking behavior and wealth. We
estimate the elasticity of ARA with respect to wealth, and use it to obtain the elasticity of RRA
with respect to wealth, based on the following expression:
ξRRA,W = ξARA,W + 1, (10)
where ξRRA,W and ξARA,W refer to the wealth elasticities of RRA and ARA, respectively. For
robustness, we also estimate the elasticity of the income-based RRA in equation (8), ξρ,W .
We exploit the panel dimension of our data and estimate these elasticities, both, in the cross
section of investors and, for a given investor, in the time series. In the cross section, wealthier
investors exhibit lower ARA and higher RRA when choosing their portfolio of loans within LC; we
refer to these elasticity estimates with the superscript xs to emphasize that they do not represent
the shape of individual preferences (i.e., ξxsARA,W , ξxsRRA,W , ξxsρ,W ). And, in the time series, investor
specific RRA increases after experiencing a negative wealth shock; that is, the preference function
exhibits decreasing RRA. The contrasting signs of the cross sectional and investor-specific wealth
elasticities indicate that preferences and wealth are not independently distributed across investors.
Below, we describe our proxies for wealth in the cross section of investors, and for wealth shocks
in the time series. Since the bulk of the analysis uses housing wealth as a proxy for investor wealth,
we focus the discussion in this section on the subsample of investors that are home-owners.33
33None of the results in this section is statistically significant in the subsample of investors that are renters. Thisis expected since housing wealth and total wealth are less likely to be correlated for renters, particularly in the timeseries. However, this is also possibly due to lack of power, since only a small fraction of the investors in our sampleare renters.
23
6.1 Cross-Sectional Evidence
We use Acxiom’s imputed net worth as of October 2007 as a proxy for wealth in the cross section
of investors. As discussed in Section 4, Acxiom’s imputed net worth is based on a proprietary
algorithm that combines names, home address, credit rating, and other data from public sources.
To account for potential measurement error in this proxy, we use a separate indicator for investor
wealth in an errors-in-variable estimation: median house price in the investor’s zip code at the time
of investment. Admittedly, house value is an imperfect indicator wealth; it does not account for
heterogeneity in mortgage level or the proportion of wealth invested in housing. Nevertheless, as
long as the measurement errors are uncorrelated across the two proxies, a plausible assumption in
our setting, the errors-in-variable estimation provides an unbiased estimate of the cross-sectional
elasticity of risk aversion to wealth.
We begin by exploring non-parametrically the relationship between the risk aversion estimates
and our two wealth proxies for the cross section of home-owner investors in our sample. Figure 4
plots a kernel-weighted local polynomial smoothing of the risk aversion measure. The horizontal
axis measures the (log) net worth and the (log) median house price in the investor’s zip code at the
time of the portfolio choice. ARA is decreasing in both wealth proxies, while income-based RRA
is increasing.
Turning to parametric evidence, we estimate the cross sectional elasticity of ARA to wealth
using the following regression:
ln (ARAi) = β0 + β1 ln (NetWorthi) + ωi. (11)
The left hand side variable is investor i’s average (log) ARA, obtained by averaging the ARA
estimates recovered from the investor’s portfolio choices during our sample period. The right-
hand side variable is investor i’s imputed net worth. Thus, the estimated β1 corresponds to the
cross-sectional wealth elasticity of ARA, ξxsARA,W .
To account for measurement error in our wealth proxy we estimate specification (11) in an
errors-in-variables model by instrumenting imputed net worth with the average (log) house value
in the zip code of residence of investor i during the sample period. Since the instrument varies
only at the zip code level, in the estimation we allow the standard errors in specification (11) to be
24
clustered by zip code. The errors-in-variables approach works in our setting because risk preferences
are obtained independently from wealth. If, for example, risk aversion were estimated from the
share of risky and riskless assets in the investor portfolio, this estimate would inherit the errors in
the wealth measure. As a result, any observed correlation between risk aversion and wealth could
be spuriously driven by measurement errors. This is not a concern in our exercise.
Table 6 shows the estimated cross sectional elasticities with OLS and the errors-in-variables
model. Our preferred estimates from the errors-in-variables model indicate that the elasticity of
ARA to wealth in the cross section is -0.074 and statistically significant at the 1% confidence level
(column 2). The non-parametric relationship is confirmed: wealthier investors exhibit a lower
ARA. The OLS elasticity estimate is biased towards zero. This attenuation bias is consistent with
classical measurement error in the wealth proxy.
The estimated ARA elasticity and equation (10) imply that the wealth-based RRA elasticity
to wealth is positive, ξxsRRA,W = 0.93. Columns 3 and 4 show the result of estimating specification
(11) using the income-based RRA as the dependent variable. The income-based RRA increases
with investor wealth in the cross section, and the point estimate, 0.078, is also significant at the
1% level (column 4). The sign of the estimated elasticity coincides with that implied by the ARA
elasticity. Overall, the results consistently indicate that the RRA is larger for wealthier investors
in the cross section.
6.2 Within-Investor Estimates
The above elasticity, obtained from the variation of risk aversion and wealth in the cross section,
can be taken to represent the form of the utility function of the representative investor only under
strong assumptions. Namely, when the distributions of wealth and preferences in the population
are independent.34 To identify the functional form of individual risk preferences we estimate the
ARA elasticity using within-investor time series variation in wealth.
House values dropped sharply during our sample period.35 Since housing represents a substan-
tial fraction of household wealth in the U.S., this decline implied an important negative wealth
34Chiappori and Paiella (2011) formally prove that any within-investor elasticity of risk aversion to wealth can besupported in the cross section by appropriately picking such joint distribution.
35In this subsample the average zip code house price declines 4% between October 2007 and April 2008. In addition,the time series house price variation is heterogeneous across investors: the median house price decline is 3.67%.
25
shock for home-owners.36 We use this source of variation, to estimate the wealth elasticity of
investor-specific risk aversion in the subsample of home-owners that invest in LC:
ln (ARAit) = αi + β2 ln (HouseV alueit) + t+ ωit. (12)
The left-hand side variable is the estimated ARA for investor i in month t. The right-hand side
variable of interest is the (log) median house value of the investor’s zip code during the month
the risk aversion estimate was obtained (i.e., the month the investment in LC takes places). The
right-hand side of specification (12) includes a full set of investor dummies as controls. These
investor fixed effects (FE) account for all cross sectional differences in risk aversion levels. Thus,
the elasticity β2 recovers the sensitivity of ARA to investor-specific shocks to wealth. We also
include a time-trend to absorb the evolution of housing prices, common to all investors, during the
period under analysis.
By construction, the parameter β2 can be estimated only for the subsample of investors that
choose an LC portfolio more than once in our sample period. Although the average number of
portfolio choices per investor is 2.0, the median investor chooses only once during our analysis
period. This implies that the data over which we obtain the within investor estimates using (12)
comes from approximately half of the original sample. To insure that the results below are rep-
resentative for the full investor sample, we also show the results of estimating specification (12)
without the investor FE to corroborate that the conclusions of the previous section are unchanged
when estimated on the subsample of investors that chose portfolios more than once and controlling
for time-trends.37
Table 7 reports the parameter estimates of specification (12), before and after including the
investor FE. The FE results represent our estimated wealth elasticities of ARA, ξARA,W . The sign
of the estimated within-investor elasticity of ARA to wealth (column 2) is the same as in the cross
section: absolute risk aversion is decreasing in investor wealth.
Equation (10) and the estimated wealth elasticity of ARA imply a negative wealth-based RRA
36According to the Survey of Consumer Finances of 2007, the value of the primary residence accounts for approxi-mately 32% of total assets for the median U.S. family (see Bucks et al., 2009).
37The estimates of ARA and the investment amount are statistically indistinguishable between those that investonce versus those that invest more than once (comparing only the first investment for those that invest more thanonce).
26
to wealth changes for a given investor, ξRRA,W , of -1.85. Column 4 report the result of estimating
specification (12) using the income-based RRA as the dependent variable. The point estimate,
-4.35, also implies a negative relationship between this alternative measure of RRA and wealth.
These results consistently suggest that investors’ utility function exhibits decreasing relative risk
aversion.
The drop in house value is an incomplete measure of the change in the investor overall wealth.
It is important, then, to analyze the potential estimation bias introduced by this measurement
error. Classical measurement error would imply that the point estimate is biased towards zero;
this estimate is therefore a lower bound (in absolute value) for the actual wealth elasticity of risk
aversion. The (absolute value) of the elasticity could be overestimated if the percentage decline
in house values underestimates the change in the investor’s total wealth. However, for error in
measurement to account for the sign of the elasticity, the overall change in wealth has to be three
times larger than the percentage drop in house value.38 This is unlikely in our setting since stock
prices dropped 10% and investments in bonds had a positive yield during our sample period.39
Therefore, even if measurement error biases the numerical estimate, it is unlikely to affect our
conclusions regarding the shape of the utility function. Finally, conditioning on investors that
invest more than once in LC may introduce selection bias. If investors with large increments in risk
aversion stop investing in LC, our computations would underestimate the effect of wealth shocks
on risk aversion. If, on the other hand, investors with large negative wealth stop investing, our
results would overestimate the wealth elasticity of risk aversion.
The observed positive relationship between investor RRA and wealth in the cross section from
the previous section changes sign once one accounts for investor preference heterogeneity. The
comparison of the estimates with and without investor FE in Table 7 confirms it. Moreover, we
show in on-line appendix C.1 that the estimated elasticities of risk aversion to wealth, both in the
cross-section of investors and for the same investor, are consistent with the observed relationship
between the total investment amount in LC and wealth.40
38We estimate the elasticity of ARA with respect to changes in house value to be –2.84. Let W be overall wealthand H be house value, then: ξARA,W = d lnARA
d lnW= −2.84 · d lnH
d lnW. The wealth elasticity of RRA is positive only if
ξARA,W > −1, which requires d lnWd lnH
> 2.84.39Between October 1, 2007 and April 30, 2008 the S&P 500 Index dropped 10% and the performance of U.S.
investment grade bond market was positive —Barclays Capital U.S. Aggregate Index increased approximately 2%.40Since the ARA and elasticity estimates do not use information total investment in LC, this consistency test
constitute an independent validation of our conclusions on investors’ risk taking behavior.
27
This implies that investors preferences and wealth are not independently distributed in the cross
section. Investors with different wealth levels may have different preferences, for example, because
more risk averse individuals made investment choices that made them wealthier. Alternatively, an
unobserved investor characteristic, such as having more educated parents, may cause an investor
the be wealthier and to be more risk averse. The results indicate that characterizing empirically
the shape of the utility function requires, first, accounting for such heterogeneity.
7 Conclusion
In this paper we estimate risk preference parameters and their elasticity to wealth based on the
actual financial decisions of a panel of U.S. investors participating in a person-to-person lending
platform. The average absolute risk aversion in our sample is 0.037. We also measure the relative
risk aversion based on the income generated by investing in LC (income-based RRA). We find a
large degree of heterogeneity, with an average income-based RRA of 2.81 and a median of 1.61.
These findings are similar to those obtained in experimental studies in the field and laboratories;
they provide an external validation in a real life investment environment to the estimates obtained
from experiments. We show that, since our estimates of risk aversion refer to the local curvature
of preferences over changes in income, the parameters estimated here do not depend on a specific
utility function and correctly describe agents’ preferences in different behavioral models.
We exploit the panel dimension of our data and estimate the elasticity of ARA and RRA with
respect to wealth, both, in the cross section of investors and, for a given investor, in the time
series. In the cross section, wealthier investors exhibit lower ARA and higher RRA when choosing
their portfolio of loans within LC. For a given investor, the RRA increases after experiencing a
negative wealth shock; that is, the average investor’s preference function exhibits decreasing RRA.
The contrasting signs of the cross sectional and investor-specific wealth elasticities indicate that
investors’ preferences and wealth are not independently distributed in the cross section. Therefore,
to empirically characterize the shape of the utility function, one needs to take the properties of the
joint distribution of preferences and wealth into account.
28
References
Arellano, M. and Bonhomme, S. (2012), ‘Identifying Distributional Characteristics in RandomCoefficients Panel Data Models’, Review of Economic Studies 79(3), 987–1020.
Barberis, N. and Huang, M. (2001), ‘Mental Accounting, Loss Aversion, and Individual StockReturns’, The Journal of Finance 56(4), 1247–1292.
Barberis, N., Huang, M. and Thaler, R. H. (2006), ‘Individual preferences, monetary gambles,and stock market participation: A case for narrow framing’, The American Economic Review96(4), 1069–1090.
Barseghyan, L., Molinari, F., O’Donoghue, T. and Teitelbaum, J. (2013), ‘The Nature of RiskPreferences: Evidence from Insurance Choices’, The American Economic Review 103(6), 2499–2529.
Barseghyan, L., Prince, J. and Teitelbaum, J. (2011), ‘Are Risk Preferences Stable Across Contexts?Evidence from Insurance Data’, The American Economic Review 101(2), 591–631.
Barsky, R. B., Juster, F. T., Kimball, M. S. and Shapiro, M. D. (1997), ‘Preference Parametersand Behavioral Heterogeneity: An Experimental Approach in the Health and Retirement Study’,The Quarterly Journal of Economics 112(2), 537–579.
Blake, D. (1996), ‘Efficiency, Risk Aversion and Portfolio Insurance: An Analysis of Financial AssetPortfolio Held by Inverstors in the United Kingdom’, The Economic Journal 106(136), 1175–1192.
Blume, M. and Friend, I. (1975), ‘The Demand for Risky Assets’, The American Economic Review65(5), 900–922.
Bombardini, M. and Trebbi, F. (2012), ‘Risk Aversion and Expected Utility Theory: an Experimentwith Large and Small Stakes’, Journal of the European Economic Association (6), 1348–1399.
Brunnermeier, M. and Nagel, S. (2008), ‘Do Wealth Fluctuations Generate Time-Varying RiskAversion? Micro-evidence on Individuals’, The American Economic Review 98(3), 713–36.
Bucks, B., Kennickel, A., Mach, T. and Moore, K. (2009), ‘Changes in U.S. Family Financesfrom 2004 to 2007: Evidence from the Survey of Consumer Finances’, Federal Reserve Bulletin95, A1–A55.
Buraschi, A. and Jiltsov, A. (2007), ‘Habit Formation and Macroeconomic Models of the TermStructure of Interest Rates’, The Journal of Finance 62, 3009–3063.
Calvet, L. E., Campbell, J. Y. and Sodini, P. (2009), ‘Fight or Flight? Portfolio Rebalancing byIndividual Investors’, The Quarterly Journal of Economics 124(1), 301–348.
Calvet, L. and Sodini, P. (2014), ‘Twin Picks: Disentangling the Determinants of Risk-Taking inHousehold Portfolios’, Journal of Finance 69(2), 867–906.
Campbell, J. and Cochrane, J. (1999), ‘By Force of Habit: A Consumption Based Explanation ofAggregate Stock Market Behavior’, Journal of Political Economy 107, 205–251.
29
Chen, K. Y. and Plott, C. (1998), ‘Non-linear Behavior in Sealed Bid First Price Auctions’, Gamesand Economic Behavior 25(1), 34–78.
Chiappori, P.-A., Gandhi, A., Salanie, B. and Salanie, F. (2008), ‘You Are What You Bet: ElicitingRisk Attitudes from Horse Races’, Unpublished Working Paper. Columbia University .
Chiappori, P.-A., Gandhi, A., Salanie, B. and Salanie, F. (2009), ‘Identifying Preferences underRisk from Discrete Choices’, American Economic Review 99(2), 356 – 362.
Chiappori, P.-A. and Paiella, M. (2011), ‘Risk Aversion is Constant: Evidence from Panel Data’,Journal of the European Economic Association 9(6), 1021–1052.
Choi, S., Fisman, R., Gale, D. and Kariv, S. (2007), ‘Consistency and Heterogeneity of IndividualBehavior under Uncertainty’, The American Economic Review 97(5), 1921–1938.
Cicchetti, C. and Dubin, J. (1994), ‘A Microeconomic Analysis of Risk Aversion and the Decisionto Self-Insure’, Journal of Political Economy 102(11), 169–186.
Cohen, A. and Einav, L. (2007), ‘Estimating Risk Preferences from Deductible Choice’, The Amer-ican Economic Review 97(3), 745–788.
Cohn, R. A., Lewellen, W. G., Lease, R. C. and Schlarbaum, G. G. (1975), ‘Individual InvestorRisk Aversion and Investment Portfolio Composition’, The Journal of Finance 30(2), 605–620.
Cox, J. and Sadiraj, V. (2006), ‘Small- and Large-stakes Risk Aversion: Implications of ConcavityCalibration for Decision Theory’, Games and Economic Behavior 56, 45–60.
Einav, L., Finkelstein, A., Pascu, I. and Cullen, M. R. (2012), ‘How General are Risk Preferences?Choice under Uncertainty in Different Domains’, The American Economic Review 21(1), 415–448.
Goeree, J. and Holt, C. (2004), ‘A Model of Noisy Introspection’, Games and Economic Behavior46(2), 365–382.
Goeree, J., Holt, C. and Palfrey, T. (2002), ‘Quantal Response Equilibrium and Overbidding inPrivate-Value Auctions’, Journal of Economic Theory 104(1), 247–272.
Goeree, J., Holt, C. and Palfrey, T. (2003), ‘Risk Averse Behavior in Generalized Matching PenniesGames’, Games and Economic Behavior 45(1), 97–113.
Gollier, C. (2001), ‘Wealth Inequality and Asset Pricing’, Review of Economic Studies 68(1), 181–203.
Gomes, F. and Michaelides, A. (2008), ‘Asset Pricing with Limited Risk Sharing and HeterogeneousAgents’, The Review of Financial Studies 21(1), 415–448.
Guiso, L. and Paiella, M. (2008), ‘Risk Aversion, Wealth, and Background Risk’, Journal of theEuropean Economic Association 6(6), 1109–1150.
Guvenen, F. (2009), ‘A Parsimonious Macroeconomic Model for Asset Pricing’, Econometrica77(6), 1711–1750.
30
Harrison, G., Lau, M. and Rutstrom, E. (2007), ‘Estimating Risk Attitudes in Denmark: A FieldExperiment’, Scandinavian Journal of Economics 109(2), 341–368.
Harrison, G., Lau, M. and Towe, C. (2007), ‘Naturally Occurring Preferences and ExogenousLaboratory Experiments: A Case Study of Risk Aversion’, Econometrica 75(2), 433–458.
Holt, C. A. and Laury, S. K. (2002), ‘Risk Aversion and Incentive Effects’, The American EconomicReview 92(5), 1644–1655.
Iyer, R., Khwaja, A., Luttmer, E. and Shue, K. (2014), ‘Screening Peers Softly: Inferring theQuality of Small Borrowers’, Unpublished Working Paper. University of Chicago .
Jullien, B. and Salanie, B. (2000), ‘Estimating Preferences Under Risk: The Case of RacetrackBettors’, The Journal of Political Economy 108(3), 503–530.
Jullien, B. and Salanie, B. (2008), Empirical Evidence on the Preferences of Racetrack Bettors, inD. Hausch and B. Ziemba, eds, ‘Efficiency of Sports and Lottery Markets’, pp. 27–49.
Kocherlakota, N. (1996), ‘The Equity Premium: It’s Still a Puzzle’, Journal of Economic Literature34(1), 42–71.
Korniotis, G. (2008), ‘Habit Formation, Incomplete Markets, and the Significance of Regional Riskfor Expected Returns’, The Review of Financial Studies 21(5), 2139–2172.
Korniotis, G. and Kumar, A. (2011), ‘Do Older Investors Make Better Investment Decisions?’,Review of Economics and Statistics 93(1), 244–265.
Lin, M., Prabhala, N. and Viswanathan, S. (2013), ‘Judging Borrowers by the Company They Keep:Friendship Networks and Information Asymmetry in Online Peer-to-Peer Lending’, ManagementScience 59(1).
Marom, D. and Sade, O. (2013), ‘Are the Life and Death of an Early Stage Venture Indeed in thePower of the Tongue? Lessons from Online Crowdfunding Pitches’, Unpublished. Working PaperHebrew University .
Menzly, L., Santos, T. and Veronesi, P. (2004), ‘Understanding Predictability’, The Journal ofPolitical Economy 112(1), 1–47.
Morin, R.-A. and Suarez, A. F. (1983), ‘Risk Aversion Revisited’, The Journal of Finance38(4), 1201–1216.
Polkovnichenko, V. (2007), ‘Life Cycle Portfolio Choice with Additive Habit Formation Preferencesand Uninsurable Income Risk’, The Review of Financial Studies 20(1), 83–124.
Post, T., van den Assem, M. J., Baltussen, G. and Thaler, R. H. (2008), ‘Deal or No Deal? DecisionMaking under Risk in a Large-Payoff Game Show’, American Economic Review 98(1), 38–71.
Rabin, M. (2000), ‘Risk Aversion and Expected-Utility Theory: A Calibration Theorem’, Econo-metrica 68(5), 1281–1292.
31
Rabin, M. and Thaler, R. (2001), ‘Anomalies: Risk Aversion’, The Journal of Economic Perspec-tives 15(1), 219–232.
Rabin, M. and Thaler, R. (2002), ‘Response from Matthew Rabin and Richard H. Thaler’, TheJournal of Economic Perspectives 16(2), 229–230.
Ravina, E. (2012), ‘Love & loans: The effect of beauty and personal characteristics in creditmarkets’, Mimeo, Columbia GSB .
Santos, T. and Veronesi, P. (2010), ‘Habit Formation, the Cross Section of Stock Returns and theCash Flow Risk Puzzle’, The Journal of Financial Economics 98(2), 385–413.
Tanaka, T., Camerer, C. F. and Nguyen, Q. (2010), ‘Risk and Time Preferences: Linking Experi-mental and Household Survey Data from Vietnam’, American Economic Review 100(1), 557–571.
Treynor, J. and Black, F. (1973), ‘How to Use Security Analysis to Improve Portfolio Selection’,Journal of Business 46(1), 66–86.
Yogo, M. (2008), ‘Prices Under Habit Formation and Reference-Dependent Preferences’, Journalof Business and Economic Statistics 26(2), 131–143.
32
Figure 1: Portfolio Tool Screen Examples for a $100 Investment
39
Figure 1: Portfolio Tool Screen Examples for a $100 Investment A. Screen 1: Interest rate – Normalized Variance “Slider”
B. Screen 2: Suggested Portfolio Summary
The website provides an optimization tool that suggests the efficient portfolio of loans for the investor's preferred risk return trade-off, under the assumption that loans are uncorrelated with each other and with outside investment opportunity. The risk measure is the variance of the diversified portfolio divided by the variance of a single investment in the riskiest loan available (as a result it is normalized to be between zero and one). Once a portfolio has been formed, the investor is shown the loan composition of her portfolio on a new screen that shows each individual loan (panel B). In this screen the investor can change the amount allocated to each loan, drop them altogether, or add others.
Note: The website provides an optimization tool that suggests the efficient portfolio of loans for the investor’s
preferred risk return trade-off, under the assumption that loans are uncorrelated with each other and with
outside investment opportunity. The risk measure is the variance of the diversified portfolio divided by the
variance of a single investment in the riskiest loan available (as a result it is normalized to be between zero
and one). Once a portfolio has been formed, the investor is shown the loan composition of her portfolio on
a new screen that shows each individual loan (panel B). In this screen the investor can change the amount
allocated to each loan, drop them altogether, or add others.
33
Figure 2: Examples of Risk Return Choices and Estimated ARA
(a) Example Baseline
41
Figure 3: Examples of Risk Return Choices and Estimated RRA
B2
B1
B3
B4
A4
C1
1.09
1.1
1.11
1.12
.2 .3 .4 .5 .6WxVrz_actual
ERz Fitted values
Theta = 1.082, ARA = .0661
B4
B2
B3
A5
B1
A4
A3
1.08
1.08
51.
091.
095
1.1
1.10
5
1 1.5 2 2.5 3WxVrz_actual
ERz Fitted values
Theta = 1.068, ARA = .0120
C4
A3
C3
E2
F2
G1
A4
C5
F5
E1
D3
G5
E3
D5
1.08
1.1
1.12
1.14
1.16
1.18
0 .5 1 1.5 2WxVrz_actual
ERz Fitted values
Theta = 1.080, ARA = .0488
G3
A2
D2D1
C5
E1
B3
A1
B1
A4
F1E4
C4
B4
D4
A5B2
B5
F2
C1
D5
E3
C2
G2
C3
1.06
1.08
1.1
1.12
1.14
0 10 20 30 40WxVrz_actual
ERz Fitted values
Theta = 1.079, ARA = .0013
Each plot represents one investment in our sample. The plotted points represent the risk and weighted return of each of the buckets that compose the investment. The dots are labeled with the corresponding risk classification of the bucket. The vertical axis measures the expected return of a risk bucket, and the horizontal axis measures the bucket variance weighted by the total investment in that bucket. The slope of the linear fit is our estimate of the absolute risk aversion (ARA). The intersection of this linear fit with the vertical axis is our estimate for the risk premium (Theta).
(b) Example Baseline
41
Figure 3: Examples of Risk Return Choices and Estimated RRA
B2
B1
B3
B4
A4
C1
1.09
1.1
1.11
1.12
.2 .3 .4 .5 .6WxVrz_actual
ERz Fitted values
Theta = 1.082, ARA = .0661
B4
B2
B3
A5
B1
A4
A3
1.08
1.08
51.
091.
095
1.1
1.10
5
1 1.5 2 2.5 3WxVrz_actual
ERz Fitted values
Theta = 1.068, ARA = .0120
C4
A3
C3
E2
F2
G1
A4
C5
F5
E1
D3
G5
E3
D5
1.08
1.1
1.12
1.14
1.16
1.18
0 .5 1 1.5 2WxVrz_actual
ERz Fitted values
Theta = 1.080, ARA = .0488
G3
A2
D2D1
C5
E1
B3
A1
B1
A4
F1E4
C4
B4
D4
A5B2
B5
F2
C1
D5
E3
C2
G2
C3
1.06
1.08
1.1
1.12
1.14
0 10 20 30 40WxVrz_actual
ERz Fitted values
Theta = 1.079, ARA = .0013
Each plot represents one investment in our sample. The plotted points represent the risk and weighted return of each of the buckets that compose the investment. The dots are labeled with the corresponding risk classification of the bucket. The vertical axis measures the expected return of a risk bucket, and the horizontal axis measures the bucket variance weighted by the total investment in that bucket. The slope of the linear fit is our estimate of the absolute risk aversion (ARA). The intersection of this linear fit with the vertical axis is our estimate for the risk premium (Theta).
(c) Automatic Buckets
42
Figure 4: Example of Risk Aversion Estimation Using Automatic and Non-Automatic Buckets for the Same Investment
A. Automatic Buckets
A2
C1
G4
F4
D2
E3
1.0
81
.11
.12
1.1
41
.16
0 .5 1 1.5 2WxVrz_actual
ERz Fitted values
Theta = 1.075, ARA = 0.048
B. Non-Automatic Buckets
D3
B2
C5
A5
B5
E4
E2
G3
1.0
81
.11
.12
1.1
41
.16
0 .5 1 1.5 2WxVrz_actual
ERz Fitted values
Theta = 1.068, ARA = 0.051
Both plots represent allocations to risk buckets of the same actual investment. As in Figure 3, the plotted points represent the risk and weighted return of each of the buckets that compose the investment. Panel A shows the buckets that were chosen by the portfolio tool (Automatic), and panel B shows buckets directly chosen by the investor (Non-Automatic). The slope of the linear fit represents the absolute risk aversion (ARA), and its intersection with the vertical axis represents the risk premium (Theta).
(d) Non-Automatic Buckets
42
Figure 4: Example of Risk Aversion Estimation Using Automatic and Non-Automatic Buckets for the Same Investment
A. Automatic Buckets
A2
C1
G4
F4
D2
E3
1.0
81
.11
.12
1.1
41
.16
0 .5 1 1.5 2WxVrz_actual
ERz Fitted values
Theta = 1.075, ARA = 0.048
B. Non-Automatic Buckets
D3
B2
C5
A5
B5
E4
E2
G3
1.0
81
.11
.12
1.1
41
.16
0 .5 1 1.5 2WxVrz_actual
ERz Fitted values
Theta = 1.068, ARA = 0.051
Both plots represent allocations to risk buckets of the same actual investment. As in Figure 3, the plotted points represent the risk and weighted return of each of the buckets that compose the investment. Panel A shows the buckets that were chosen by the portfolio tool (Automatic), and panel B shows buckets directly chosen by the investor (Non-Automatic). The slope of the linear fit represents the absolute risk aversion (ARA), and its intersection with the vertical axis represents the risk premium (Theta).
Note: Plots (a) and (b) represent examples of two different investments in our sample. The plotted pointsrepresent the risk and weighted return of each of the buckets that compose the investment. The dotsare labeled with the corresponding risk classification of the bucket. The vertical axis measures the expectedreturn of a risk bucket, and the horizontal axis measures the bucket variance weighted by the total investmentin that bucket. The slope of the linear fit is our estimate of the absolute risk aversion (ARA). The intersectionof this linear fit with the vertical axis is our estimate for the risk premium (θ).
Plots (c) and (b) represent allocations to risk buckets of the same actual investment. Panel (c) shows the
buckets that were chosen by the portfolio tool (Automatic), and panel (d) shows buckets directly chosen by
the investor (Non-Automatic).
34
Figure 3: Investment-by-Investment Bias Distribution
(a) ARA: Automatic and Non-Automatic Choices
010
2030
40D
ensi
ty
-.15 -.1 -.05 0 .05 .1ARA Non-Automatic - ARA Automatic
Kernel density estimateNormal density
kernel = epanechnikov, bandwidth = 0.0034
(b) θ: Automatic and Non-Automatic Choices
010
2030
40D
ensi
ty
-.1 -.05 0 .05 .1 .15Theta Non-Automatic - Theta Automatic
Kernel density estimateNormal density
kernel = epanechnikov, bandwidth = 0.0031
Note: Subsample of investments that combine buckets chosen through the optimization tool (Automatic)
with buckets chosen manually (Non-Automatic). The figures plot the distribution of the difference between
the estimates of ARA and θ obtained using Non-Automatic and Automatic buckets separately, for the same
investment. For comparison, the figures include a Normal distribution with the corresponding standard
deviation.
Figure 4: Risk Aversion and Wealth in the Cross Section
(a) ARA and Wealth
.045
.05
.055
.06
.065
ARA
10 11 12 13 14 15log Net Worth/ log Median House Prices
ARA and Net Worth ARA and House Prices95% C.I.
(b) Income-Based RRA and Wealth
2.5
33.
54
RR
A
10 11 12 13 14 15log Net Worth/ log Median House Prices
RRA and Net Worth RRA and House Prices95% C.I.
Note: Subsample of home-owners. The vertical axis plots a weighted local second degree polynomial smooth-
ing of the risk aversion measure. The observations are weighted using an Epanechnikov kernel with a band-
width of 0.75. The horizontal axis measures the (log) net worth and the (log) median house price at the
investor’s zip code at the time of the portfolio choice, our two proxies for investor wealth.
35
Table 1: Borrower and Investor Characteristics
Variable Mean Std. Dev. Median(1) (2) (3)
A. Borrower CharacteristicsFICO score 694.3 38.2 688.0Debt to Income 0.128 0.076 0.128Monthly Income ($) 5,428 5,963 4,250Amount borrowed ($) 9,224 6,038 8,000
B. Investor CharacteristicsMale 83% 100%Age 43.4 15.0 40.0Married 56% 100%Home Owner 75% 100%Net Worth, Imputed ($1,000) 663.0 994.4 375.0Median House Value in Zip Code ($1,000) 385.1 285.8 292.4% Change in House Price, 10-2007 to 04-2008 -4.08% 4.97% -3.67%
Note: October 2007 to April 2008. FICO scores and debt to income ratios are recovered from each borrower’scredit report. Monthly incomes are self reported during the loan application process. Amount borrowed isthe final amount obtained through Lending Club. Lending Club obtains investor demographics and networth data through a third party marketing firm (Acxiom). Acxiom uses a proprietary algorithm to recovergender from the investor’s name, and matches investor names, home addresses, and credit history detailsto available public records to recover age, marital status, home ownership status, and net worth. We useinvestor zip codes to match the LC data with real estate price data from the Zillow Home Value Index. TheZillow Index for a given geographical area is the median property value in that area.
36
Table 2: Descriptive Statistics
Sample/Subsample: All Investments Diversified Investments With Real Estate Data(1) (2) (3)
Mean S.D Median Mean S.D Median Mean S.D Median
A. Unit of observation: investor-bucket-month(N = 50,254) (N = 43,662) (N=38,980)
Investment ($) 302.8 2251.4 50.0 86.0 206.9 50.0 88.4 215.1 50.0N Projects in Bucket 1.9 1.8 1.0 2.0 1.8 1.0 2.0 1.8 1.0Interest Rate 12.89% 2.98% 12.92% 12.91% 2.96% 12.92% 12.92% 2.96% 12.92%Default Rate 2.77% 1.45% 2.69% 2.78% 1.45% 2.84% 2.79% 1.45% 2.84%E(PV $1 investment) 1.122 0.027 1.122 1.122 0.027 1.123 1.122 0.027 1.123Var(PV $1 investment) 0.036 0.020 0.035 0.027 0.020 0.022 0.027 0.020 0.022
B. Unit of observation: investor-month(N = 5,191) (N = 3,745) (N=3,294)
Investment 2,932 28,402 375 1,003 2,736 375 1,044 2,864 400N Buckets 9.7 8.7 7.0 11.7 8.4 10.0 11.8 8.5 10.0N Projects 18.8 28.0 8.0 23.3 28.9 14.0 23.7 29.3 14.0E(PV $1 investment) 1.121 0.023 1.121 1.121 0.021 1.121 1.121 0.021 1.121Var(PV $1 investment) 0.012 0.016 0.005 0.005 0.007 0.003 0.005 0.006 0.002
Note: Each observation in panel A represents an investment allocation, with at least 2 risk buckets, by investor i in risk bucket z in montht. In panel B, each observation represents a portfolio choice by investor i in month t. An investment constitutes a dollar amount allocationto projects (requested loans), classified in 35 risk buckets, within a calendar month. Loan requests are assigned to risk buckets according tothe amount of the loan, the FICO score, and other borrower characteristics. Lending Club assigns and reports the interest rate and defaultprobability for all projects in a bucket. The expectation and variance of the present value of $1 investment in a risk bucket is calculatedassuming a geometric distribution for the idiosyncratic monthly survival probability of the individual loans and independence across loanswithin a bucket. The sample in column 2 excludes portfolio choices in a single bucket and non-diversified investments. The sample in column3 also excludes portfolio choices made by investors located in zip codes that are not covered by the Zillow Index.
37
Table 3: Unconditional distribution of estimated risk aversion parameters
ARA θExpected IncomeIncome Based RRA
(1) (2) (3) (4)
mean 0.03706 1.086 125.6 2.81sd 0.02450 0.028 327.3 3.54
p1 -0.00812 1.045 4.1 -0.16p10 0.01177 1.059 8.2 0.29p25 0.02293 1.076 16.0 0.56p50 0.04404 1.086 45.5 1.61p75 0.04813 1.094 107.8 3.59p90 0.05300 1.104 292.1 7.16p99 0.08675 1.156 1,226.4 16.79
N 3,286 3,286 3,286 3,286
Note: Absolute Risk Aversion (ARA) and intercept θ obtained through the OLS estimation of the followingrelationship for each investment:
E [Rz] = θi +ARAi · Wixizniz
· σ2z + ξiz
where the left (right) hand side variable is expected return (idiosyncratic variance times the investmentamount) of the investment in bucket z. The income based Relative Risk Aversion (RRA) is the estimatedARA times the total expected income from the investment in Lending Club. pN represents the N th percentileof the distribution.
38
Table 4: Mean risk aversion and systematic risk premium by month
ARA θExpected IncomeIncome Based RRA
(1) (2) (3) (4)
2007m10 0.029 1.057 173.411 1.292(0.020) (0.013) (603.8) (1.054)
2007m11 0.033 1.064 110.913 1.208(0.018) (0.013) (333.3) (0.946)
2007m12 0.037 1.066 77.858 1.442(0.016) (0.013) (193.3) (1.506)
2008m1 0.036 1.083 171.576 2.790(0.031) (0.038) (496.4) (3.636)
2008m2 0.040 1.089 116.564 3.123(0.022) (0.017) (275.2) (3.791)
2008m3 0.037 1.097 139.392 3.752(0.025) (0.026) (269.2) (4.197)
2008m4 0.039 1.089 61.579 1.932(0.027) (0.023) (107.3) (2.209)
Note: Absolute Risk Aversion (ARA) and intercept θ obtained through the OLS estimation of the followingrelationship for each investment:
E [Rz] = θi +ARAi · Wixizniz
· σ2z + ξiz
where the left (right) hand side variable is expected return (idiosyncratic variance times the investmentamount) of the investment in bucket z. The income based Relative Risk Aversion (RRA) is the estimatedARA times the total expected income from the investment in Lending Club. Standard deviations in paren-thesis.
39
Table 5: Estimates from Automatic and Non-Automatic Buckets
ARA θ
Mean S.D. Mean S.D.(1) (2) (3) (4)
A. Full Sample (n = 243)Automatic 0.0373 0.0213 1.079 0.0205Non-Automatic 0.0360 0.0194 1.080 0.0221Automatic - Non-Automatic -0.0014 0.0202 0.001 0.0209
B. Subsample: October-December 2007 (n = 76)Automatic 0.0357 0.0232 1.061 0.0134Non-Automatic 0.0336 0.0190 1.063 0.0191Automatic - Non-Automatic -0.0021 0.0221 0.001 0.0167
C. Subsample: January-April 2008 (n = 167)Automatic 0.0381 0.0204 1.086 0.0183Non-Automatic 0.0371 0.0195 1.088 0.0187Automatic - Non-Automatic -0.0010 0.0194 0.001 0.0226
Note: Descriptive statistics of the Absolute Risk Aversion (ARA) and θ obtained as in Table 3, over thesubsample of investments where the estimates can be obtained separately using Automatic (buckets suggestedby optimization tool) and Non-Automatic (buckets chosen directly by investor) bucket choices for the sameinvestment. The table reports the mean and standard deviation of both estimates and of the difference forthe same investment, for the full sample and for 2007 and 2008 separately.
40
Table 6: Risk Aversion and Wealth, Cross Section Estimates
Dependent Variable ARA Income Based RRA log(Net Worth)
(in logs) OLS Errors-in-Var OLS Errors-in-Var First Stage(1) (2) (3) (4) (5)
log(Net Worth) -0.010** -0.074*** 0.020** 0.078***(0.004) (0.021) (0.008) (0.029)
log(House Value) 1.327***(0.120)
R2 0.003 0.004 0.068Observations (investors) 1,794 1,794 1,791 1,791 1,794
Note: Estimated elasticity of risk aversion to wealth in the cross section. Columns 1 and 3 present the OLSestimation of the between model and columns 2 and 4 present the errors-in-variables estimation using themedian house value in the investor’s zip code as an instrument for net worth. The dependent variables arethe (log) absolute risk aversion (column 1 and 2) and income-based relative risk aversion (columns 3 and4), averaged for each investor i across all portfolio choices in our sample. The right hand side variable isthe investor (log) net worth (from Acxiom). Column 5 reports the first stage of the instrumental variableregression: the dependent variable is (log) net worth and the right hand side variable is the average (log)median house price in the investor’s zip code (from Zillow). Standard errors are heteroskedasticity robustand clustered at the zip code level. *, **, and *** indicate significance at the 10%, 5%, and 1% levels ofconfidence, respectively.
41
Table 7: Risk Aversion and Wealth Shocks, Investor-Specific Estimates
Dep. Variable ARA Income Based RRA
(in logs) Pooled OLS Investor FE Pooled OLS Investor FE(1) (2) (3) (4)
log(House Value) -0.132*** -2.847* 0.158*** -4.351***(0.042) (1.478) (0.052) (1.640)
Investor Fixed Effects No Yes No YesTime Trend Yes Yes Yes YesR2 0.023 0.021 0.027 0.010Observations 1,843 1,843 1,843 1,843Investors 1,041 1,041 1,041 1,041
Note:Estimated investor-specific elasticity of risk aversion to wealth. The left hand side variables are the(log) absolute risk aversion (columns 1 and 2) and income-based relative risk aversion (columns 3 and 4),obtained for investor i for a portfolio choice in month t. The right hand side variables are the (log) medianhouse price in the investor’s zip code in time t, and an investor fixed effect (omitted). Standard errors areheteroskedasticity robust and clustered at the zip code level. *, **, and *** indicate significance at the 10%,5%, and 1% levels of confidence, respectively.
42
top related